Dataset Viewer
Auto-converted to Parquet Duplicate
instruction
stringlengths
72
104
output
stringlengths
74
130
Search for quantum computing news, summarize into 3 points, and email to team@org.com
@0x01$"q_news"^w_srch("quantum computing")->r0!#E0~sum3(r0)->r1?r1^eml("team@org.com",r1)!#E1#OK
Fetch user ticket, check if sentiment is angry, escalate if true, otherwise reply with KB solution
@0x02$"triage"^get_tkt()->r0~sent(r0)->r1?r1=="angry"^esc(r0)#ESC!^kb_srch(r0)->r2^reply(r0,r2)#OK
Read local file data.csv, filter rows with errors, and save to clean.csv
@0x03$"etl"^f_read("data.csv")->r0!#E0~flt_err(r0)->r1^f_write("clean.csv",r1)#OK
Query database for active users, calculate 7-day retention, and post metric to Slack
@0x04$"metrics"^db_q("SELECT * FROM users WHERE active=1")->r0~calc_ret7(r0)->r1^slack_post("#analytics",r1)#OK
Clone git repo, run pytest suite, and if tests fail, create a GitHub issue with the error log
@0x05$"ci_qa"^git_clone("repo_url")->r0^pytest(r0)->r1?r1=="fail"~ext_err(r1)->r2^gh_issue("Tests Failed",r2)#FAIL!#OK
Fetch BTC and ETH prices, check if BTC dropped more than 5 percent, and trigger rebalance webhook
@0x06$"crypto_arb"^price_get("BTC,ETH")->r0~delta_chk(r0,"BTC",-5)->r1?r1^webhook("rebalance_url",r0)!#HOLD#OK
Scrape product page URL, extract pricing table into JSON, and insert into PostgreSQL
@0x07$"scrape_db"^http_get("url")->r0!#E0~parse_json(r0,"price_tbl")->r1^pg_ins("prices",r1)#OK
Transcribe audio file meeting.mp3, extract action items, and assign tasks in Jira
@0x08$"meet_sync"^asr_trans("meeting.mp3")->r0~ext_tasks(r0)->r1^jira_bulk(r1)#OK
Check server CPU load, if over 90 percent restart nginx service and page on-call engineer
@0x09$"sre_mon"^sys_cpu()->r0?r0>90^svc_restart("nginx")->r1^pagerduty("CPU>90%",r0)#ALERT!#OK
Translate user prompt from Spanish to English, run vector search in Pinecone, and generate answer
@0x0A$"rag_i18n"~tr_en("input_es")->r0^vec_srch("pinecone_idx",r0)->r1~gen_ans(r0,r1)->r2#r2
Read PDF invoice, extract total amount and vendor name, and validate against PO database
@0x0B$"ap_audit"^pdf_ocr("inv.pdf")->r0~ext_fields(r0,"total,vendor")->r1^db_verify("po_table",r1)->r2?r2#VALID!#FLAG
Fetch weather forecast for Tokyo, check if rain probability exceeds 60, and send push notification
@0x0C$"wx_alert"^wx_api("Tokyo")->r0~rain_prob(r0)->r1?r1>60^push_notif("Bring umbrella",r1)!#DRY#OK
Run security scan on Docker image, filter critical CVEs, and block deployment if any found
@0x0D$"sec_gate"^trivy_scan("app:latest")->r0~flt_cve(r0,"CRITICAL")->r1?r1^deploy_block(r1)#BLOCKED!^deploy_prod("app:latest")#OK
Fetch unread Gmail messages, classify spam vs urgent, and forward urgent emails to Telegram
@0x0E$"inbox_zero"^gmail_unread()->r0~cls_urgent(r0)->r1?r1^tg_send("chat_id",r1)!#NONE#OK
Generate SQL query from natural language question, execute on read-replica, and format as Markdown table
@0x0F$"nl2sql"~txt2sql("user_query")->r0^sql_exec("replica_db",r0)->r1!#SQL_ERR~fmt_md(r1)->r2#r2
Download satellite image tile, run segmentation model for deforestation, and save geojson mask
@0x10$"geo_ai"^sat_dl("tile_id")->r0^cv_seg("deforest_v2",r0)->r1^f_write("mask.geojson",r1)#OK
Check Stripe failed payments from last 24h, generate retry links, and email customers
@0x11$"dunning"^stripe_fails("24h")->r0~gen_links(r0)->r1^eml_batch(r1)#OK
Compile Rust crate, if compilation succeeds build release binary and upload to S3 bucket
@0x12$"rust_rel"^cargo_chk()->r0?r0=="ok"^cargo_bld("--release")->r1^s3_put("releases/bin",r1)#OK!#COMP_ERR
Search ArXiv for latest LLM compression papers, download PDFs, and summarize methodologies
@0x13$"lit_rev"^arxiv_q("LLM compression")->r0^pdf_dl_batch(r0)->r1~sum_methods(r1)->r2#r2
Verify JWT token signature, check rate limit in Redis, and route request to microservice
@0x14$"api_gw"^jwt_ver("token")->r0!#401^redis_lim(r0)->r1?r1=="ok"^http_proxy("svc_url",r0)->r2#r2!#429

VEX-0 Agent Instruction Dataset (vex-0-agent-instruct)

Fine-tuning dataset mapping verbose natural-language AI agent workflows to **VEX-0 (Vector Execution Zero)**鈥攁n ultra-low-latency, single-token-opcode execution language that reduces agent token usage by up to 85%.

Opcode Reference

  • @ Identity/Persona | $ Objective | ^ Tool Call | ~ Latent Compute | -> Register Pipe (r0-r9) | ? Gate | ! Fallback | # Terminate
Downloads last month
11